Parametric Statistical Generalization of Uniform-Hazard Earthquake

Ground Motions
Bryce W. Dickinson, M.ASCE1; and Henri P. Gavin, M.ASCE2

Abstract: Sets of ground-motion records used for seismic hazard analyses typically have intensity measures corresponding to a particular
hazard level for a site (perhaps conditioned on a particular intensity value and hazard). In many cases the number of available ground
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

motions that match required spectral ordinates and other criteria (such as duration, fault rupture characteristics, and epicentral distance)
may not be sufficient for high-resolution seismic hazard analysis. In such cases it is advantageous to generate additional ground motions
using a parameterized statistical model calibrated to records of the smaller data set. This study presents a statistical parametric analysis of
ground-motion data sets that are classified according to a seismic hazard level and a geographic region and that have been used
extensively for structural response and seismic hazard analyses. Parameters represent near-fault effects such as pulse velocity and pulse
period, far-field effects such as velocity amplitude and power-spectral attributes, and envelope characteristics. A systematic fitting of
parameterized pulse functions to the individual ground-motion records, of parameterized envelopes to individual instantaneous ground-
motion amplitudes, and of parameterized power-spectral density functions to averaged power spectra result in probability distributions for
ground-motion parameters representative of particular seismic hazard levels for specific geographical regions. This methodology presents
a means to characterize the variability in a set of ground-motion records of physically meaningful parameters. DOI:
10.1061/(ASCE)ST.1943-541X.0000330. © 2011 American Society of Civil Engineers.

CE Database subject headings: Ground motion; Parameters; Statistics; Earthquakes.

Author keywords: Synthetic ground motion; Ground-motion parameter; Near-fault effect.

Introduction Engineering, Vol. 137, No. 3, March 1, 2011. ©ASCE, ISSN 0733-9445/
2011/3-410–422/$25.00.
Earthquake engineers and engineering seismologists have describing the earthquake source, the path of the wave
appreciated the importance of uncertainty in earthquake ground propagation, and the local and regional geophysical conditions
motions for more than 60 years (Housner 1947). Earthquake through empirical relationships (Power et al. 2008). Stochastic
ground-motion data sets, even when scaled to a common intensity models are employed to describe and reproduce the stochastic
measure, exhibit significant variability in spectral and temporal characteristics of a set of observed ground motions without
characteristics (Housner 1947; Cornell 1968; Conte and Peng necessarily considering the geophysical properties of the event
1997). Despite the growing availability of ground-motion records, (Shinozuka and Deodatis 1988). Researchers have further
ground-motion characteristics having an intensity with a certain developed methods to combine attenuation and stochastic process
return period at a site can be known only approximately because models (Boore 2005b).
of natural variability in earthquake source mechanisms, seismic Probabilistic seismic hazard analysis (PSHA) combines local
wave propagation, and local site effects. Consequently, seismic hazard models, attenuation models, and site effects to
expectations about the potential ground-motion characteristics at determine the spectral accelerations and other ground-motion
a site are suitably interpreted on a probabilistic basis. characteristics of a given return period at a site (Petersen et al.
Procedures to generalize and simulate strong ground motions 2008). Performance-based design of structures with nonlinear
may be generally divided into three main classifications: behavior, on the other hand, requires realistic ground motions
computational geophysical models, empirical attenuation models, representative of the hazard designated by PSHA (McGuire
and empirical stochastic models. Computational geophysics 1995). Considering the dispersion in intensity measures, even
employs finite-difference orfinite- when conditioned on an intensity measure with a specific hazard
elementmethodstonumericallysolve elastodynamic wave level, a large number of records may be necessary for a high-
equations to simulate wave propagation through a modeled media resolution performance-based design.
(Graves 1998). Attenuation models attempt to relate peak ground To this end, a general approach toward the statistical
motions and spectral response values to parameters characterization of ground motions from a sample of
representative groundmotion records is presented. Parameter
1 values that describe the stochastic content, the long-period
Design Enginner, Rutherford & Chekene, 55 Second St., Suite 600, velocity pulse, and the timemodulating envelope function are
San Francisco, CA 94105.
2 determined individually for each record. For this study sets of
Assistant Professor, Dept. Civil and Environmental Engineering, ground-motion records are used that have been widely
Duke Univ., Box 90287, Durham, NC 27708. appropriated and which remain in use by many researchers. This
Note. This manuscript was submitted on June 30, 2008; approved on methodology may be applied as easily to other sets of ground
September 26, 2010; published online on October 21, 2010. Discussion motions.
period open until August 1, 2011; separate discussions must be submitted
for individual papers. This paper is part of the Journal of Structural

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 410

J. Struct. Eng. 2011.137:410-422.

 Background, Literature Review, and Motivation throughout a geophysical region (Graves 1998; Bielak et al. 2005;
Olsen et al. 2008).
Methods for simulating earthquake ground motions typically Increased awareness of the destructive capability of near-fault
describe strong ground motions as nonstationary stochastic velocity pulses has motivated several studies examining the
processes. response of structures and their contents subjected to this type of
A number of methods have been proposed for generalizing and forcing (Bertero et al. 1978; Rao and Jangid 2001; Zhang and
simulating the nonstationary behavior of earthquake ground Iwan 2002a, b; Mavroeidis and Papageorgiou 2003; Mavroeidis et
motions (Grigoriu 1995; Conte and Peng 1997; Michaelov et al. al. 2004; Tian et al. 2007). The energy wave released upon the
1999; Shinozuka et al. 1999; Iyengar 2001; Mukherjee and Gupta slip of a fault can exhibit a number of characteristics including
2002; Wen and Gu 2004; Gu and Wen 2007). The more common forward directivity, permanent translation, and radiation effects
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

of these stochastic process descriptors involve time-dependent (Mavroeidis and Papageorgiou 2003). When the fault rupture
spectral parameters and a time-modulating envelope function. The velocity is near the shear wave velocity, seismic energy
wellknown Kanai-Tajimi ground-motion spectrum (Kanai 1957; accumulates at the wave front as the rupture propagates; ground
Tajimi 1960) models ground-motion as the response of a damped velocity records at sites that are close to the rupture and located
linear oscillator excited by broadband acceleration. The power- ahead of the slip exhibit a distinct high-amplitude pulse in the
spectral density function for this model contains two physically fault-normal component. Such pulses are typically observed near
meaningful parameters: ground frequency and damping. To the beginning of the record (Somerville 2003; Tothong et al.
achieve nonstationarity, such a stationary stochastic process may 2007). Mavroeidis and Papageorgiou have modeled near-fault
be multiplied by a deterministic envelope function (Jennings et al. earthquake groundmotion records containing long-period velocity
1968; Iyengar and Iyengar 1969; Liu 1970; Lin and Yong 1987). pulses (Mavroeidis and Papageorgiou 2003). This model is
Shinozuka and Deodatis comprehensively reviewed stochastic parameterized by the period, the amplitude, the number of cycles,
ground-motion simulation methodologies including the filtering and the phase of the pulse (all of which have a clear physical
of Gaussian white noise, the filtering of Poisson processes, the interpretation). The ultimate generation of synthetic near-fault
spectral representation of stochastic processes, and simulation ground motions is accomplished through the superposition of a
based on stochastic wave theory (Shinozuka and Deodatis 1988). coherent, long-period velocity pulse with incoherent seismic
A companion review paper (Kozin 1988) discusses autoregressive radiation. A number of subsequent studies have examined the
moving average (ARMA) models to parametrically match target mechanisms and modeling of near-fault ground motions to study
ground motions. Many of these models efficiently simulate the the resulting structural response (Somerville 2003; Bray and
evolving amplitude and frequency content of a recorded ground Rodriguez-Marek 2004; Tian et al. 2007). A quantitative measure
motion and contain parameters that depend on the records chosen of pulselike characteristics determined that 91 of the 3,551
for model calibration. The point source stochastic simulation records in the Pacific Earthquake Engineering Research Center-
model is a commonly used numerical procedure for simulating Next Generation Attenuation project database contains a pulse
ground-motion records (Boore 2003, 2005b). This method makes (Baker 2007). The destructive potential of ground motions
use of a deterministic envelope function applied to a filtered white containing a strong velocity pulse indicates a need to include a
noise process, in which the power-spectral density and duration of pulse simulation model in stochastic process models for the
the record are related empirically to the earthquake fault rupture, simulation of demanding ground motions.
path effects, and site conditions. This method has been shown to Stochastic methods for the simulation of ground-motion
be effective in simulating the stochastic content of ground records must therefore contain several key characteristics. The
motions (Boore 2003) and has bridged a gap between geophysical process model must be capable of simulating the nonstationarity
attenuation relationships and stochastic process models. of strong ground motions. It must be capable of the inclusion of
Rather than attempt to model the range of ground motions that both long-period velocity pulses, as well as incoherent radiation
a given structure may experience over its lifetime, critical effects. Ideally, the model will be simple with a small number of
excitation methods have been introduced (Takewaki 2005), which physically meaningful parameters. Current methods in simulating
assume there is a single worst-case input excitation that will cause ground motions that include a near-fault velocity pulse and
the most significant response in a given structure. This critical stochastic content have not yet considered the probability
excitation is based on the properties of the elastic or inelastic distributions of the model parameters or the relation of these
structural system using the input energy and the energy input rate probability distributions to particular seismic hazard levels or
as the measures of criticality. With this method, a building may be geographical regions.
designed to withstand the worst case input, which is certainly The present study aims to quantify the statistical variation of
important for critical structures. This method is not amenable to ground-motion parameters fit to samples of ground-motion
estimating exceedance rates or other probabilistic measures. records and to generate additional ground motions containing
Computational geophysical models enable simulation of these same statistical characteristics. The ground-motion model
ground motions with characteristics that may be underrepresented involves the superposition of a long-period velocity pulse and a
in catalogs of recorded ground motions and associated empirical nonstationary stochastic acceleration record. Several investigators
models. Such characteristics include forward directivity effects, have assembled samples of ground-motion records representative
residual ground displacement, basin effects, and fault-zone of a given hazard level as designated by a PSHA (Iervolino and
channeling. These methods involve the coupling of source models Cornell 2005; McGuire 1995; Somerville et al. 1997; Tothong and
with wave propagation models to simulate ground-motion activity Baker

411 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

 2007). In this study, models are calibrated to the ground-motion
data sets developed for the SAC steel project (Somerville et al.
1997), a joint venture of the Structural Engineers Association of
California, the Applied Technology Council, and the California
Universities for Research in Earthquake Engineering. The five
data sets explored in this study were developed to represent the
hazard probabilities of exceedance of 2% in 50 years and 10% in
50 years in Los Angeles and Seattle. A set containing 40 records
with near-fault effects is also explored. These data sets have been
regarded as statistical samples for the potential ground-motion for
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

a given hazard level and specific geographical region criteria. As

a result, the SAC ground motions have been used widely for
seismic analysis and performance-based design (Chang et al.
2002; Lee and Foutch 2002; Wen and Song 2003; Yun et al.
2002; Morgan and Mahin 2008).
Statistical analysis of the model parameters for each of the
near-fault, Los Angeles, and Seattle data sets allow for waveform
parameters to be drawn from distributions with means, variances,
and parameter correlations calibrated to samples of historic
ground-motion records. In a companion study (Gavin and
Dickinson 2011), correlations between input model parameters
and spectral responses are plotted as “correlation spectra.” These
spectra are used to reduce the number of input parameters to only
those that associate with variability in structural response.

Ground-Motion Parameterization
In this study, synthetic earthquake ground motions are generated
through the superposition of a stochastic ground velocity record
with a single coherent long-period velocity pulse. The stochastic
ground-motion velocity record is generated by first simulating an
enveloped and unscaled stochastic ground acceleration record,
auðtÞ

auðtiÞ ¼ ^eðti;τ0;τ1;τ2;τ3Þ XNþ1½S^ðf k;f g;ζgÞ1=2

k¼1

× cosð2πf kti þ θkÞ ð1Þ

where ^Sðf k;f g;ζgÞ = power spectrum of the stochastic content of
the acceleration record parameterized by ground frequency, f g,
and a ground damping, ζg

^Sðf k;f g;ζgÞ ¼ ð 1 ðf

k=ðf2gζÞ2gÞf2k=þ ðf gÞ22ζgf k=f gÞ2 ð2Þ

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 412

J. Struct. Eng. 2011.137:410-422.

 ððti τ0Þ=τ1Þ2 ^eðti;τ0;τ1;τ2;τ3Þ ¼ for ττ00 þ≤ tτ1≤≤τ0t þ≤ττ10 þ τ1 þ τ2
3
as shown in Fig. 2. 8< 1 for π2 t T 2
ð Þ
ϕ pk

s
ðÞ¼
vuðtÞðVs=maxjvuðtÞjÞ, where Vs is a parameter denoting the peak
as shown in Fig. 3. This function is essentially a single Gabor
of the stochastic ground velocity record. The artificial ground
wavelet in which the scale and shift parameters are continuous
velocity record is then found by combining the scaled stochastic
ground velocity with a velocity pulse and have a clear physical interpretation. The artificial ground
acceleration record agðtÞ is found by differentiating v gðtÞ from
frequencies of 0 to f hi in the frequency domain.
^vðtiÞ ¼ vsðtiÞ þ vpðtiÞ ð4Þ In summary, 12 individual parameter values are used in this
study to completely characterize a given ground-motion record:
pulse amplitude, Vp; pulse period, Tp; number of pulse cycles, Nc;
The pulse model has five parameters: the peak pulse velocity, V p; pulse arrival time, Tpk; pulse phase, ϕ; peak velocity of the
the period of the pulse, T p; the number of cycles in the pulse, N c; stochastic content, Vs; central ground frequency, f g; ground
the location of the pulse, Tpk; and the phase of the pulse, ϕ damping, ζg; time of no ground acceleration, τ0; time to peak

: exp½ðti τ0 τ1 τ2Þ=τ3 for τ0 þ τ1 þ τ2 ≤ t

flated fromlo An unscaled stochastic ground velocity recordand f hi ain the frequency-domain (uðtÞ by integrating frequency
components betweenBoore 2005avuð;tÞDickinsonis calcu-pðti;Vp;Tp;Nc;Tpk; Þ ¼ V× cosp exp2π t4TTp pk
NcTpϕ ð5Þ

2008). The scaled stochastic ground velocity record is v t

1 τ1 τ2
1
0.8
Spectral power

0.8
Envelope

0.6
2ζ g fg 0.6 τ3
0.4 0.4

0.2 0.2

0 0
0 2 4 6 8 10 0 2 4 6 8 10 12 14
fg
Frequency, ©
413 / JOURNAL OF STRUCTURAL ENGINEERING HzASCE / MARCH 2011 Time, s
Fig. 1. Power spectrum parameters Fig. 2. Envelope function parameters
J. Struct. Eng. 2011.137:410-422.
as shown in Fig. 1, and ^eðt;τ0;τ1;τ2;τ3Þ is an envelope function.
For digital simulation, discrete frequencies f k are uniformly
spaced from f lo to f hi; f k ¼ f lo þ ðk 1Þðf hi f loÞ=N. The phase
angle at each frequency, θk, is a uniformly distributed random
variable between 0 and 2π (Shinozuka and Jan 1972; Shinozuka
and Deodatis 1991). In this study, N ¼ 1024 and the frequency
limits, f lo and f hi, depend on the type of ground motion to be
simulated. The envelope function used here separates the ground
motion into four time periods (Jennings et al. 1968): a time of no
ground acceleration, τ0; the time the ground motion takes to reach
its peak acceleration, τ1; the time the acceleration is at the plateau
peak region, τ2; and an exponential decay time constant, τ3

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 414

J. Struct. Eng. 2011.137:410-422.

 acceleration record was bandpass filtered between 0.1 and 1.5 Hz
and integrated once to give the velocity record and integrated
again to give the displacement record for the long-period content
of the ground motion. These long-period acceleration, velocity,
and displacement records were visually examined to determine
the presence of a coherent pulse. Pulses are typically visible in the
velocity record and are even more apparent in the displacement

Parameter Vp (cm=s) Tp (s) Nc Tpk (s) ϕ (rad) fg (Hz) ζg τ0 (s) τ1 (s) τ2 (s) τ3 (s)
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

Minimum 5 0.5 0.5 0.1 0.5 0.05 0.1 0.1 0.1 0.1
ππ
Maximum 500 9.5 5.0 20 10 3.0 50 50 50 50
record. Numerical values for the peak pulse velocity, V p, the
period of the pulse, Tp, the number of cycles in the pulse, Nc, the
location of the pulse, Tpk, and the phase of the pulse, ϕ, were
identified such that the pulse Eq. (5) roughly matched the long-
period and high-amplitude content of the velocity record. These
values served as the initial guess for a nonlinear least squares fit
of the pulse equation to the filtered velocity record

χ2ðVp;Tp;Nc;Tpk;ϕÞ ¼ 1 Xm ½vðtiÞ vpðti;Vp;Tp;Nc;Tpk;ϕÞ2

2
6
i¼1 ð Þ

Parameter constraints were enforced during the fitting process to

Tpk ensure that parameter estimates remained within physically
50
motivated bounds, as shown in Table 1.
40
Nonlinear least squares methods sometimes converge to local
Vp T pcN
30
minimums and occasionally different initial guesses for the pulse
parameters resulted in different parameter estimates. In such
Velocity, cm/s

20
10
cases, decisions were made as to which estimate was most
0
reasonable.
−10
−20 Table 1. Parameter Estimate Constraints
−30
Individual pulse records were designated as “easy,” “medium,” or
−40 Tp “difficult” to fit, depending on the ease of finding a pulse, the
−50
0 2 4 8 10 12 14 quality of the fit, the speed of convergence, and the difficulty of
Time,s the choices made in the process.
An “easy” to fit record exhibits a single distinct velocity pulse:
Fig. 3. Pulse parameters for ϕ¼ 0 a clear peak or set of peaks in the record that is longer in period
and higher amplitude than any other surrounding peak. The pulse
is often the main oscillatory component in the low-frequency
acceleration, τ1; duration of plateau region, τ2; and acceleration bandpass filtered data set, though some records contain additional
decay time constant, τ3. higher frequency content. An easy pulse to fit typically consists of
approximately one cycle, different initial guesses converge to the
same parameter estimates within a few iterations, and the modeled
Parameter Estimation pulse displacements are close to the actual displacement record
and are occasionally detectable in the acceleration record as well.
This section describes the estimation of the 12 ground-motion Examples of records with easy to fit pulses to are shown in Fig. 4.
parameters for five of the ground-motion data sets developed for Records that contained both a long-period low-amplitude pulse
the SAC steel project. Estimates of parameter values and their and a short-period high-amplitude pulse were fit with medium
confidence intervals were computed using the Levenberg- difficulty. Sometimes an extremely high-amplitude peak
Marquardt method in three related nonlinear least squares (approximately 150% or higher than other peaks in the record)
problems. would be selected as the pulse over some slightly longer period
The pulse parameters were estimated first. To extract but significantly lower amplitude peaks. Conversely, sometimes a
longperiod data from the ground-motion records, each very long-period pulse would be selected over other peaks with

415 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

much shorter periods and not significantly larger amplitudes. If
additional reasoning is needed in picking a peak for the pulse fit,
the pulse closer to the beginning of the record was chosen.
Usually, once a choice is made as to which peak constituted the
desired pulse, the resulting fit would converge to a desirable result
within a few iterations. Another reason a fit might be labeled
medium in difficulty would be because of the need to constrain
the parameter estimates to sensible values. For example, the
number of pulse cycles, Nc, was constrained to a minimum value
of 0.5. The nonlinear least squares procedure for a few of the
records attempted to drive the number of cycles below this value,
however, to compensate for this, the pulse amplitude value, V p,
would grow arbitrarily large, leading to nonsensical results.
Changes in parameter values throughout the convergence process
were carefully examined for consistency before choosing a final
result. Examples of records with medium pulse-fitting difficulty
from each of the groundmotion suites are shown in Fig. 5.
A record labeled as “difficult” to fit was typically one for
which the least squares method would have trouble converging or
would converge to different results based on different initial
guesses with no clear distinction between pulse options. In some
cases, particularly in the Seattle records, it was so difficult to
isolate a distinct pulse that it was decided that no discernible pulse
was present in the record. In such cases, the pulse velocity would
be assigned a value of zero. Several other records contained what
appeared to be multiple cycle pulses where convergence was
difficult because of either the variability of results from different
initial guesses or simply the inability of the converged result to
capture the pulse behavior. In most cases, either convergence (or
at least general stability in parameters) was achieved that would
capture most of any pulselike characteristic of the record based on
choices that aimed to fit longperiod and larger amplitude trends
toward the beginning of the record. The greatest problem with the
difficult pulse fits was that the pulse parameters were outliers; the
method would frequently sacrifice capturing the high-amplitude
oscillations for longer-period

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 416

J. Struct. Eng. 2011.137:410-422.

Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Easy pulse fits for each data set (solid line is fit, dashed line is data): (a) near-fault, nf11; (b) LA 10%=50 yr, la03; (c) LA 2%=50 yr, la30;
(d) SE 10%=50 yr, se14; (e) SE 2%=50 yr, se23

trends. Examples of records with difficult to fit pulses are shown e ti

ð Þ¼
in Fig. 6. The numbers of “easy,”“medium,” difficult,” and “no
a t
pulse” records for each of the ground-motion suites are shown in q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið iÞ2 þ
Table 2. ~a ti 7
ð Þ2ffi ð Þ
To find Vs, the peak of the stochastic part of the velocity
record, the acceleration time history was bandpass filtered
between 0.1 and 20 Hz and integrated once. The velocity pulse fit The envelope function, Eq. (3), was fit to the instantaneous
was then subtracted from the overall velocity record to leave only acceleration amplitudes normalized to a peak value of 1. The
the stochastic part of the velocity record. The peak of the
stochastic velocity record, Vs, was then determined and recorded. envelope fit function was divided by p2 so that the approximation
To fit the envelope, the instantaneous acceleration amplitude, would pass through the normalized data. In fitting the envelope
eand its Hilbert transform,ðtiÞ, was computed from the original
function, theffiffi residuals were weighted by the square root of
acceleration record,~aðtiÞ aðtiÞ,
the data such that the higher amplitude values of the normalized
acceleration would be more significant χ2ðτ0;τ1;τ2;τ3Þ

¼ 21 Xi¼m1 eðtiÞ=maxðeðtiÞÞ
1=p^eðeffiffiffiffiffiffiffitið;tτiÞffi0;τ1;τ2;τ3Þ=p2ffiffi2 ð8Þ

417 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

It was found that this weighting method resulted in envelope fits
that reached their maximum value very close to the time of the
peak ground acceleration (within the first half of the ground-
motion record), which was considered preferable to envelopes
that had maxima toward the end of the record. Subtracting the
pulse acceleration before fitting the envelope did not noticeably
affect envelope parameters. Initial guesses for the four envelope
parameters were chosen by visual inspection. Despite a variety of
initial guesses with long values of τ2, the least squares method
consistently removed the
plateauregion,indicatingthatthisparametercouldbeexcludedfrom
the envelope function. Examples of envelope functions from each
of the ground-motion suites are shown in Fig. 7.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 418

J. Struct. Eng. 2011.137:410-422.

Fig. 5. Medium pulse fits for each data set (solid line is fit, dashed line is data): (a) near-fault, nf18; (b) LA 10%=50 yr, la07; (c) LA 2%=50 yr,
419 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011
la34;
(d) SE 10%=50 yr, se13; (e) SE 2%=50 yr, se26
J. Struct. Eng. 2011.137:410-422.
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

The power-spectral density (PSD) of the difference between converged within a few iterations. The PSD fits for each of the
each acceleration record and its pulse was estimated using ground-motion suites are shown in Fig. 8.
Welch’s averaged periodogram method and normalized to a Parameter Distributions
maximum of 1. The normalized spectra were then averaged over
all records for each of the five data sets to obtain an averaged The systematic fitting of waveform expressions to the SAC
PSD for each of the five sets. Each of the five average PSDs were groundmotion data sets provides a statistical sample of the
then fit with Eq. (2). In fitting power spectra, the residuals were characteristics of each suite. The near-fault set provides 40
weighted by the square root of Sðf iÞ, such that the higher sample measurements of 10 of the model parameters, and the
amplitude values of the power spectrum were considered more others provide 20 sample measurements of 10 of the model
significant parameters. The remaining two parameters, ground frequency, f g,
and ground damping, ζg, are obtained from spectra averaged over
each set, thereby giving one value per set. The means and
variances of these statistical samples are provided in Table 3. In
χ2ðf g;ζgÞ ¼ 12 Xi¼m1 Sðf iÞ this table, the mean values for ground frequency, f g, and ground
1=p^SðffiffiffiffiffiffiffiffiSfði;ffiÞgffi ;ζgÞ2 ð9Þ damping, ζg, were obtained via the nonlinear least squares fitting
of the averaged PSD functions for each data set. The reported
standard deviations for these two parameters are simply the
standard parameter errors as determined by the fitting procedure.
Although the parameter describing the initial period of no ground
This method was applied consistently to each of the five average motion, τ0, was required for fitting the envelopes, it simply
PSD functions and the resulting parameter values for the ground represents an arbitrary amount of measurement time before the
damping, ζg, and ground frequency, f g, were recorded along with strong ground motion begins. As such, this parameter has no
their associated standard errors. The averaged PSDs were easily effect in computing structural responses. Therefore, the
fit by the Levenberg-Marquardt method; parameter estimates

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 420

J. Struct. Eng. 2011.137:410-422.

distributions of values for τ0 were not explored further. Since this
time

421 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

Fig. 6. Difficult pulse fits for each data set (solid line is fit, dashed line is data): (a) near-fault, nf20; (b) LA 10% =50 yr, la19; (c) LA 2%=50 yr,
JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 422
la31;
(d) SE 10%=50 yr, se11; (e) SE 2%=50 yr, se36
J. Struct. Eng. 2011.137:410-422.
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

period is removed from models of simulated ground motions, the 10% in 50-year records. For the near-fault data set the pulse
arrival in time of the pulse, Tpk, was also adjusted to account for amplitudes are 60% larger than the amplitude of the stochastic
content. For the two Los Angeles data sets, the pulse amplitudes
this change, by subtracting τ0 from Tpk.
are 33% larger than the amplitude of the stochastic content. In the
The parameter estimates show that the 2% in 50-year Los two Seattle data sets the pulse amplitudes are only half of the
Angeles data set has the largest velocity amplitudes, the near-fault stochastic content amplitudes. The pulses identified from the
data set has the second-largest amplitude parameters. The mean Seattle data set are therefore relatively inconsequential. The
velocity amplitudes for the Los Angeles 2% in 50-year records are coefficient of variation (COV) in the pulse velocity is
a little more than twice the mean velocity amplitudes for the Los approximately 70% for near-fault records and approximately 50%
Angeles 10% in 50-year records. The mean velocity amplitudes for the other Los Angeles records. The stochastic velocity
for the Seattle 2% in 50-year records are about three times larger amplitude’s COV is about 50% for all data sets.
than the mean velocity amplitudes for the Seattle Mean pulse periods ranged from 2.58 to 2.85 s across the five
data sets. The pulse period’s COV is about 60% for all data sets.
Based on these waveforms, pulse period statistics are relatively
Table 2. Pulse Fit Difficulty for Five Data Sets
insensitive to the hazard level or geographical location.
LA2 LA10 SE2 SE10 The central frequency of the stochastic content of Seattle
Data set Near-fault in50 in50 in50 in50
ground motions (≈1:20 Hz) is higher than that of Los Angeles
Easy 21 9 10 8 5
ground motions (≈0:75 Hz), and the duration of Seattle ground
Medium 12 8 8 5 6
motions is about three times those of Los Angeles. In all five data
Difficult 7 3 2 0 1
No pulse 0 0 0 7 8
sets the bandwidth (ζf g) of the stochastic content is about 1:0=s.

423 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

 Fig. 7. Example envelope fits for each data set (solid line is fit, dashed line is data): (a) near-fault, nf02; (b) LA 10% =50 yr, la16; (c) LA 2%=50
yr, la34; (d) SE 10%=50 yr, se13; (e) SE 2%=50 yr, se39

Cumulative distributions for each of the nine ground-motion Summary and Conclusions
parameters for the five data sets are shown in Figs. 9–13. The
associated standard errors for every parameter estimate are The generation of synthetic ground motions in this study is
displayed as horizontal error bars. The standard error bars shown accomplished via the superposition of a coherent velocity pulse
are truncated by the parameter value bounds provided in Table 1. with a stochastic acceleration record corresponding to a specific
Several parameters have physically motivated minimum values, PSD function and multiplied by a time-modulating
as shown in the last row of Table 3. These offsets are subtracted envelope function. Estimation of 12 ground-motion waveform
from each of the sample values prior to the estimation of the mean parameters was accomplished via a systematic nonlinear least
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

and variance of each distribution. These offset values must then squares analysis for 120 of the SAC ground-motion records. Five
be added to the randomly generated parameter value prior to the parameters describe the pulse: pulse amplitude, V p; pulse period,
ground-motion simulation. For example, the number of cycles in a Tp; number of cycles, Nc; pulse arrival time, Tpk; and pulse phase,
pulse, Nc, for records in the “la10in50” set, can be considered as ϕ. One parameter describes the peak of the stochastic content of
0.5 plus a lognormal random variable with a mean of 0.8 and a the velocity, Vs. Four parameters describe the envelope: the time
standard deviation of 0.8. This procedure serves the purpose of before ground motion begins, τ0; the period of increasing
generating a better fit of the cumulative distribution function as amplitude, τ1; a plateau region of strong ground motion, τ2; and a
well as ensuring that physically meaningful parameter minimums period of slow amplitude decay, τ3. Two parameters describe the
are enforced during simulations. Empirical cumulative PSD function: the ground frequency, f g; and the ground damping,
distributions of the parameter estimates match lognormal and ζg. The ground frequency and ground damping parameters were fit
gamma distribution functions. The difference between the for the average PSD for each data set rather than for individual
distributions is indiscernible in their upper tails. records because power spectra computed from short-duration
transient waveforms have a
1 (a) Near Fault

0.8
Power spectrum

0.6

0.4

0.2

0 5 10 15
Frequency, Hz

1 (b) LA2in50 1 (c) LA10in50

0.8 0.8
Power spectrum

Power spectrum

0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 5 10 15 0 5 10 15
Frequency, Hz Frequency, Hz

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 424

J. Struct. Eng. 2011.137:410-422.

 1 (d) SE2in50 1 (e) SE10in50

0.8 0.8

Power spectrum

Power spectrum
0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 5 10 15 0 5 10 15
Frequency, Hz Frequency, Hz
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Power spectra fits for each data set (solid line is fit, dashed line is data): (a) near-fault; (b) LA 10%=50 yr; (c) LA 2%=50 yr; (d) SE
10%=50 yr; (e) SE 2%=50 yr

Table 3. Fit Statistics and Offsets for Five Data Sets

Parameter

Suite Vp (cm=s) Tp (s) Nc (—) Tpk (s) ϕ (rad) Vs (cm=s) τ1 (s) τ2 (s) τ3 (s) fg (Hz) ζg —

nrfault mean 90.83 1.78 0.88 5.04 2.72 57.04 4.02 0.48 4.08 0.76 1.41
std.dev. 74.10 1.06 0.66 3.23 1.64 33.06 3.66 0.75 2.31 0.08 0.14

la10in50 mean 45.28 2.05 0.80 4.97 2.71 34.83 3.77 0.14 4.48 0.78 1.76
std.dev. 22.81 1.25 0.82 3.32 1.42 16.77 2.69 0.14 3.10 0.08 0.14

la2in50 mean 99.99 1.92 0.64 4.81 3.43 74.88 4.46 0.53 3.81 0.73 1.36
std.dev. 47.62 0.94 0.47 3.49 1.47 31.72 3.64 0.77 1.94 0.07 0.14

se10in50 mean 11.60 1.95 0.84 6.11 3.26 22.22 12.25 0.14 10.40 1.33 0.82
std.dev. 11.97 0.65 0.30 2.47 1.50 11.91 13.04 0.10 8.28 0.14 0.08

se2in50 mean 34.13 2.02 1.17 6.61 3.73 63.76 15.66 0.13 10.62 1.15 0.81
std.dev. 36.68 1.05 0.59 3.37 1.34 26.12 14.63 0.14 7.24 0.14 0.08

offset 0 0.8 0.5 0 π 10 1 0 1 0 0

425 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

 1 1 1
0.8 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
0 0

50100 150 200 250 300

Vp, cm/s

0.8

0.6

0.4

0.2

2 4 0
Tpk, s
φ, rad

1
0.8
0.6
0.4
0.2
0
2 4

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 426

J. Struct. Eng. 2011.137:410-422.

 Fig. 9. Parameter estimates, standard errors, and cumulative distributions of ground-motion parameter estimates for the SAC “nrfault” set (large
dot is parameter estimates, horizontal line is standard parameter errors, dotted line is gamma distribution, solid line is lognormal distribution): (a)
Vp; (b) Tp; (c) Nc; (d) Tpk; (e) ϕ; (f) Vs; (g) τ1; (h) τ2; (i) τ3
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Parameter estimates, standard errors, and cumulative distributions of ground-motion parameter estimates for the SAC “la10in50” set (large
dot is parameter estimates, horizontal line is standard parameter errors, dotted line is gamma distribution, solid line is lognormal distribution): (a)
Vp; (b) Tp; (c) Nc; (d) Tpk; (e) ϕ; (f) Vs; (g) τ1; (h) τ2; (i) τ3

427 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

 1 1 1
0.8 0 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
0 0
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

Vp, cm/s

0.8

0.6

0.4

0.2

2 4 12 14 0
Tpk, s

1
0.8
0.6
0.4
0.2
0
2 4 12 14 16

τ1, s τ2, s τ3, s

Fig. 11. Parameter estimates, standard errors, and cumulative distributions of ground-motion parameter estimates for the SAC “la2in50” set (large
dot is parameter estimates, horizontal line is standard parameter errors, dotted line is gamma distribution, solid line is lognormal distribution): (a)
Vp; (b) Tp; (c) Nc; (d) Tpk; (e) ϕ; (f) Vs; (g) τ1; (h) τ2; (i) τ3

428 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

 τ1, s τ2, s τ3, s

Fig. 12. Parameter estimates, standard errors, and cumulative distributions of ground-motion parameter estimates for the SAC “se2in50” set (large
dot is parameter estimates, horizontal line is standard parameter errors, dotted line is gamma distribution, solid line is lognormal distribution): (a)
Vp; (b) Tp; (c) Nc; (d) Tpk; (e) ϕ; (f) Vs; (g) τ1; (h) τ2; (i) τ3

120 140

14 16

429 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.

 1 1 1
0.8 0.8 3 0.8
-2 Vr, cm/s
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
0 0
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 5 10 15

20 25 30 τ2, s τ3, s

Fig. 13. Parameter estimates, standard errors, and cumulative distributions of ground-motion parameter estimates for the SAC “se10in50” set (large
dot is parameter estimates, horizontal line is standard parameter errors, dotted line is gamma distribution, solid line is lognormal distribution): (a)
Vp; (b) Tp; (c) Nc; (d) Tpk; (e) ϕ; (f) Vs; (g) τ1; (h) τ2; (i) τ3

high variance. Sample means, variances, and distribution CMS-0402490 (NEES Operations). Any opinions, findings, and
functions were calculated for each of the other 10 parameters for conclusions, or recommendations expressed in this material are
each data set, and parameter distributions are represented well by those of the writers and do not necessarily reflect the views of the
lognormal and gamma distribution functions. National Science Foundation. The writers also express sincere
The pulse period statistics (Tp ≈ 2:7 s 60%), the bandwidth thanks to the reviewers, whose perceptive comments were
statistics (ζgf g ≈ 1:0=s 10%), and the COVof the peak stochastic instrumental in revising this manuscript.
velocity (50%) are similar across the five suites of ground motion
analyzed in this study. The central frequency of the stochastic
ground motion f g is about 1.20 Hz in the Seattle records and about References
0.75 Hz for the Los Angeles records. The duration of the Seattle Baker, J. (2007). “Quantitative classification of near-fault ground motions
records is about three times longer than the Los Angeles records, using wavelet analysis.” Bull. Seismol. Soc. Am., 97(5), 1486–1501.
and velocity pulses are relatively inconsequential in the Seattle
Bertero, V., Mahin, S., and Herrera, R. (1978). “Aseismic design
records, which are generally far-field records from subduction-
implications of near-fault San Fernando earthquake records.”
zone ruptures.
Earthquake Eng. Struct. Dyn., 6(1), 31–42.
This analysis provides a quantitative measure of the variability
Bielak, J., Ghattas, O., and Kim, E. (2005). “Parallel octree-based finite
among ground motions though a parameterized ground-motion
element method for large-scale earthquake ground-motion
waveform model. A companion study (Gavin and Dickinson
2011) uses this information to investigate correlations between simulation.” Comput. Model. Eng. Sci., 10(2), 99–112.
parameters, to identify parameters having a significant impact on Boore, D. (2003). “Simulation of ground motion using the stochastic
structural response, and to develop response-spectrum-compatible method.” Pure Appl. Geophys., 160, 635–676.
models for biaxial synthetic ground motions representative of the Boore, D. (2005a). “On pads and filters: Processing strong-motion data.”
original data sets. Provided that ground-motion parameters can be Bull. Seismol. Soc. Am., 95(2), 745–750.
identified or scaled for a sufficiently large number of hazard Boore, D. (2005b). “SMSIM-FORTRAN programs for simulating ground
levels, this simulation procedure enables high-resolution fragility motions from earthquakes: Version 2.3.” OFR 96-80-A, USGS,
analyses. Reston, VA.
Bray, J., and Rodriguez-Marek, A. (2004). “Characterization of
forwarddirectivity ground motions in the near-fault region.” Soil Dyn.
Acknowledgments Earthquake Eng., 24(11), 815–828.
This material is based upon work supported by the the Civilian Chang, S., Makris, N., Whittaker, A., and Thompson, A. (2002).
Research and Development Foundation for the Independent States “Experimental and analytical studies on the performance of hybrid
of the Former Soviet Union (CRDF) under Award No. MG1- isolation systems.” Earthquake Eng. Struct. Dyn., 31(2), 421–443.
2319-CH-02 and by the National Science Foundation under Grant Conte, J. P., and Peng, B. F. (1997). “Fully nonstationary analytical
No. NSF-CMMI-0704959 (NEES Research), and Grant No. NSF- earthquake ground-motion model.” J. Eng. Mech., 123(1), 15–24.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011 / 430

J. Struct. Eng. 2011.137:410-422.

 Cornell, C. (1968). “Engineering seismic risk analysis.” Bull. Seismol. Mukherjee, S., and Gupta, V. (2002). “Wavelet-based generation of
Soc. Am., 58(5), 1583–1606. spectrum-compatible time-histories.” Soil Dyn. Earthquake Eng., 22,
Dickinson,B.(2008).“Aparametricstatisticalgeneralizationandsimulation 799–804.
of uniform hazard earthquake ground motions.” Master’s thesis, Dept. Olsen, K., et al. (2008). “TeraShake2: Spontaneous rupture simulations of
of Civil and Environmental Engineering, Duke Univ., Durham, NC. M-w 7.7 earthquakes on the southern San Andreas fault.” Bull.
Gavin, H. P., and Dickinson, B. W. (2011). “Generation of uniform- Seismol. Soc. Am., 98(3), 1162–1185.
hazard earthquake ground motions.” J. Struct. Eng., 137(3), 423–432. Petersen, M., et al. (2008). “Documentation for the 2008 update of the
Graves, R. (1998). “Three-dimensional finite-difference modeling of the United States national seismic hazard maps.” OFR 2008-1128, USGS,
San Andreas fault: Source parameterization and ground-motion Reston, VA.
levels.” Bull. Seismol. Soc. Am., 88(4), 881–897. Power, M., Chiou, B., Abrahamson, N., Bozorgnia, Y., Shantz, T., and
Grigoriu, M. (1995). “Probabilistic models and simulation methods for Roblee, C. (2008). “An overview of the NGA project.” Earthquake
Downloaded from ascelibrary.org by Northeastern Univ Library on 07/09/13. Copyright ASCE. For personal use only; all rights reserved.

seismic ground accleration.” Meccanica, 30(1), 105–124. Spectra, 24(1), 3–21.

Gu, P., and Wen, Y. (2007). “A record-based method for the generation Rao, P., and Jangid, R. (2001). “Performance of sliding systems under
of tridirectional uniform hazard-response spectra and ground motions nearfault motions.” Nucl. Eng. Des., 203(2-3), 259–272.
using the Hilbert-Huang transform.” Bull. Seismol. Soc. Am., 97(5), Shinozuka, M., and Deodatis, G. (1988). “Stochastic process models for
1539–1556. earthquake ground motion.” Probab. Eng. Mech., 3(3), 114–123.
Housner, G. (1947). “Characteristics of strong-motion earthquakes.” Bull. Shinozuka, M., and Deodatis, G. (1991). “Stochastic wave models for
Seismol. Soc. Am., 37, 19–31. stationary and homogeneous seismic ground motion.” Struct. Saf.,
Iervolino, I., and Cornell, C. (2005). “Record selection for nonlinear 10(1–3), 235–246.
seismic analysis of structures.” Earthquake Spectra, 21(3), 685–713. Shinozuka, M., Deodatis, G., Zhang, R., and Papageorgiou, A. (1999).
Iyengar, R. (2001). Probabilistic methods in earthquake engineering, H. “Modeling, synthetics and engineering applications of strong
Aref and J. W. Phillips, eds., Kluwer Academic, The Netherlands. earthquake wave motion.” Soil Dyn. Earthquake Eng., 18(3), 209–
Iyengar, R., and Iyengar, K. (1969). “A nonstationary random process 228.
model for earthquake accelerograms.” Bull. Seismol. Soc. Am., 59(3), Shinozuka, M., and Jan, C. M. (1972). “Digital simulation of random
1163–1188. processes and its applications.” J. Sound Vibra., 25(1), 111–128.
Jennings, P., Housner, G., and Tsai, N. (1968). “Simulated earthquake Somerville, P. (2003). “Magnitude scaling of the near fault rupture
motions.” Technical Rep., Earthquake Engineering Research directivity pulse.” Phys. Earth Planet. Inter., 137, 201–212.
Laboratory, California Institute of Technology, Pasadena, CA. Somerville, P., Smith, N., Punyamurthula, S., and Sun, J. (1997).
Kanai, K. (1957). “Seismic-empirical formula for the seismic “Development of ground-motion time histories for phase 2 of the
characteristics of the ground.” Bull. Earthquake Res. Inst., Univ. FEMA/SAC steel project.” Rep. No. SAC/BD-97/04, FEMA,
Tokyo, 35, 309–325. Washington, DC.
Kozin, F. (1988). “Autoregressive moving average models of earthquake Tajimi, H. (1960). “A statistical method of determining the maximum
records.” Prob. Eng. Mech., 3(2), 58–63. response of a building structure during an earthquake.” Proc., 2nd
Lee, K., and Foutch, D. (2002). “Seismic performance evaluation of World Conf. on Earthquake Engineering, Springer, Berlin, 781–798.
preNorthridge steel frame buildings with brittle connections.” J. Takewaki, I. (2005). “A comprehensive review of seismic critical
Struct. Eng., 128(4), 546–555. excitation methods for robust design.” Adv. Struct. Eng., 8(4), 349–
Lin, Y., and Yong, Y. (1987). “Evolutionary Kanai-Tajimi earthquake 363.
models.” J. Eng. Mech., 113(8), 1119–1137. Tian, Y., Yang, Q., and Lu, M. (2007). “Simulation method of near-fault
Liu, S. (1970). “Evolutionary power spectral density of strong-motion pulse-type ground motion.” Acta Seismologica Sinica, 20(1), 80–87.
earthquakes.” Bull. Seismol. Soc. Am., 60(3), 891–900. Tothong, P., Cornell, C., and Baker, J. (2007). “Explicit directivity-pulse
Mavroeidis, G., Dong, G., and Papageorgiou, A. (2004). “Near-fault inclusion in probabilistic seismic hazard analysis.” Earthquake
ground motions, and the response of elastic and inelastic singledegree- Spectra, 23(4), 867–891.
of-freedom (sdof) systems.” Earthquake Eng. Struct. Dyn., 33, 1023– Wen, Y., and Gu, P. (2004). “Description and simulation of nonstationary
1049. processes based on Hilbert spectra.” J. Eng. Mech., 130(8), 942–951.
Mavroeidis, G., and Papageorgiou, A. (2003). “A mathematical Wen, Y., and Song, S. (2003). “Structural reliability/redundancy under
representation of near-fault ground motions.” Bull. Seismol. Soc. earthquakes.” J. Struct. Eng., 129(1), 56–67.
Am., 93(3), 1099–1131. Yun, S., Hamburger, R., Cornell, C., and Foutch, D. (2002). “Seismic
McGuire, R. (1995). “Probabilistic seismic hazard analysis and design performance evaluation for steel moment frames.” J. Struct. Eng.,
earthquakes: Closing the loop.” Bull. Seismol. Soc. Am., 85(5), 1275– 128(4), 534–545.
1284. Zhang, Y., and Iwan, W. (2002a). “Active interaction control of tall
Michaelov, G., Sarkani, S., and Lutes, L. (1999). “Spectral characteristics buildings subjected to near-field ground motions.” J. Struct. Eng.,
of nonstationary random processes—a critical review.” Struct. Saf., 128(1), 69–79.
21(3), 223–244. Zhang, Y., and Iwan, W. (2002b). “Protecting base-isolated structures
Morgan, T., and Mahin, S. (2008). “Performance-based design of seismic from near-field ground motion by tuned interaction damper.” J. Eng.
isolated buildings considering multiple performance objectives.” Mech., 128(3), 287–295.
Smart Struct. Syst., 4(5), 655–666.

431 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / MARCH 2011

J. Struct. Eng. 2011.137:410-422.